SMARTCART

By Jun Rong, Javier and Yi Cheng

What is iT?

Time saving

Powered by machine learning

Syncing of groceries

across devices

SmartCart is an app that leverages technology to help everyone with their grocery shopping

How does

it work?

Smart

Groceries

Sharing

Enter your groceries as normal

While shopping, use the smart camera to compare prices and keep track of groceries automatically

Data is synced with you, and your family instantly

MOtivation

​

• Buying too much

​

• Duplicate purchases

​

• Difficulty in comparing prices

​

• Buying goods without comparing information

about the project

​

• SmartCart syncs
• shopping lists
• purchase information

​

• Automatically detects
• Item type
• Price
• Nutritional information (planned)

Examples

Price detection

• Uses on device OCR
• Faster response
• Does not require network

​

• Minimal increment to app size

​

​

Technologies

Machine

LEarning

The backbone: MobileNetV2

​

With the time constraint in mind, we decided to take the method of

Transfer Learning: “Transferring” one machine learning model’s knowledge into another, like a 2-for-1 deal

​

​

By retraining the MobileNetV2 model on our more

da focused dataset, we could not only save time, but

a also gain a high accuracy in our detection.

​

​

a MobileNetV2 was chosen due to our goal of having

an accurate, low-latency prediction time, with the extra extra constraint of being run on a phone.

​

​

​

​

​

Firebase

• SmartCart uses Firebase
• Firestore syncs shopping list data between devices
• Uses Firebase ML Kit for OCR and running custom models such as the grocery identifier.

​

• Firebase crashlytics
• Allows faster response to app crashes

​

Devops

• Automated builds using GitHub Actions

​

​

​

​

​

​

​

​

• APKs built automatically for fast deployment

THANK YOU!